Clean books, reconciled accounts, and a close cycle that finishes on time

The problem

Books drift quietly. A vendor charge shows up as SQ *MERCHANT 4421 on one statement and Square Inc on the next, and the auto-categorize rule misses both. Bank feeds disconnect for a weekend, transactions duplicate when the connection restores, and nobody catches it until the following month. By the time someone opens the reconciliation tab, three accounts are off by amounts that no longer correspond to any single line item.

The deeper problem is process. There is no shared definition of when a month is closed, who signed off, and what evidence supports the numbers. Receipts live in email threads, in a shared drive, in someone's phone. Department managers code expenses inconsistently because the rules live in a spreadsheet that has not been updated since the last fiscal year. When the CPA asks a question in March, reconstructing the answer takes hours instead of seconds.

This is the gap ATCS closes. Not by replacing your CPA, and not by selling a generic SaaS form. By giving the business a custom-tailored ledger, reconciliation, and close workflow that reflect the departments, locations, and accounts you actually operate.

How it works

AI-assisted categorization with chart-of-accounts learning

Every imported transaction runs through a categorization model that learns from your historical coding decisions and your specific chart of accounts. When a bookkeeper recategorizes a vendor or splits a charge across departments, the system records that decision and applies it the next time the same merchant string appears. Confidence scores surface on every suggestion, so the bookkeeper reviews edge cases and lets the routine ones flow through.

Multi-account reconciliation center with parallel workflows

Operating accounts, savings, multiple credit cards, Stripe, Square, and merchant deposit accounts each get their own reconciliation worksheet. Bookkeepers can work several accounts in parallel without losing context, and the center shows beginning balance, cleared transactions, outstanding items, and ending balance against the statement. Variance is exposed at the line level rather than buried in a single end-of-month report.

Close cycle: open, in_review, closed

Each accounting period moves through three explicit states. While the period is open, transactions can be edited and recategorized. When the bookkeeper marks the period in_review, edits require an admin override and a reason that is recorded in the audit log. Once the period is closed, the ledger is locked for that month and any subsequent adjustment becomes a journal entry with full provenance.

Audit log and document attachments per transaction

Every categorization change, reconciliation action, and close transition is captured in the audit log with the user, timestamp, and prior value. Receipts, invoices, and supporting documents attach directly to the transaction record rather than to a separate folder system. Rollup views let an owner or CPA see, for any account or department, who touched what during the period and what evidence backs each entry.

What you get

• Bank feed import with duplicate detection across operating, savings, card, and processor accounts
• CSV import for any account that does not support direct feed
• AI categorization tuned to your chart of accounts and department structure
• Vendor normalization that resolves merchant string variants to a single payee
• Encrypted vendor TIN storage for 1099 reporting
• Multi-account reconciliation center with parallel worksheets per account
• Close cycle workflow with open, in_review, and closed states and locked-period enforcement
• Per-transaction document attachments and rollup audit log
• Role-based access: owner, admin, bookkeeper, and viewer
• Month-end close measured in days, not weeks, once the workflow is established
• CPA handoff bundle including P&L, balance sheet, general ledger register, reconciliation reports, and a document index
• Department- and location-level coding that matches how the business actually reports internally

FAQ

How is this different from QuickBooks Online or Xero?

QuickBooks and Xero are configured by the customer to fit a generic template. ATCS is built around your actual departments, locations, account structure, and reconciliation cadence, with the close workflow and audit controls implemented in code rather than improvised in a shared spreadsheet. The platform is owned by the business, not rented per seat.

Can the AI categorize industry-specific accounts?

Yes. The categorization model trains on your historical ledger and chart of accounts, so industry-specific buckets such as cost of goods by SKU class, prevailing-wage labor splits, or location-specific occupancy expenses become first-class categories. Bookkeeper corrections feed back into the model, so accuracy improves with use rather than degrading.

What actually happens during a month-end close?

The bookkeeper finishes categorization, completes reconciliation for each account, and moves the period to in_review. An admin or owner reviews the variance reports, signs off, and transitions the period to closed, which locks the ledger for that month. Any later adjustment is recorded as a dated journal entry with the original transactions intact.

How do I reconcile a payment processor like Stripe or Square?

Processor accounts get their own reconciliation worksheet that ties gross charges, processor fees, refunds, chargebacks, and net deposits to the operating account. The system matches deposit batches to the underlying transactions, so the bookkeeper sees fee accruals and timing differences rather than a single net number that cannot be explained.

Is the data backed up off-site?

Yes. ATCS infrastructure includes encrypted off-site backups on a defined retention schedule, along with role-based access controls and an audit log that captures every material change. See the Infrastructure & backup page for the full backup posture.

What AI handles that bookkeepers shouldn't have to

Good bookkeepers are expensive, in short supply, and quietly burning out on the same repetitive work every month. The volume tier of bookkeeping — categorizing thousands of transactions, watching reconciliation drift across half a dozen accounts, chasing the same vendor name spelled four different ways — is where the hours disappear and the mistakes creep in. AI is genuinely better at that tier, because it doesn't get tired at row 3,000 and it doesn't forget how you coded a charge in February. The point isn't to replace the bookkeeper. It's to give them their judgment hours back.

AI does this better

• Categorize a 4,000-row month-end batch in seconds with a confidence score on each line — a bookkeeper does it across three days, and consistency drifts as fatigue sets in by Wednesday afternoon.
• Reconcile operating, savings, two credit cards, Stripe, Square, and PayPal in parallel — a bookkeeper works them serially, one tab at a time, and a Friday discrepancy waits until Monday.
• Surface stale reconciliation accounts at 2am when nobody's logged in — a bookkeeper sees them when they next open the screen, which might be after the 15th.
• Normalize vendor strings across "AMZN Mktp," "Amazon.com*1A2B3," and "AMAZON WEB SERVICES" — a bookkeeper either creates duplicate vendor records or spends an hour cleaning the list quarterly.
• Learn from corrections against your specific chart of accounts so the same fix doesn't have to be made twice — a bookkeeper relies on memory and sticky notes, and a new staffer starts the learning curve over.
• Track close-cycle state continuously across open, in_review, and closed periods, with an append-only audit log of every status change — a bookkeeper relies on a spreadsheet checklist that nobody else can audit later.
• Run read-only sweeps on demand — list_uncategorized_transactions, summarize_period, list_stale_reconciliations — without touching the books, so the bookkeeper gets a clean morning queue instead of hunting for what needs attention.
• Attach the receipt or invoice to the specific transaction at ingestion — a bookkeeper does this manually at month-end, and the documents that go missing are always the ones the auditor asks about.

What only a human bookkeeper should still do

• Decide how to split a unique transaction across departments, projects, or owners when no rule exists yet — that's a policy call, not a pattern match.
• Call the owner about an unusual charge before posting it, and read the answer for what isn't said.
• Sign off on the close before the period locks — accountability has to sit with a person.
• Handle the conversation with the CPA at year-end about reclassifications, accruals, and judgment items the books can't infer.
• Investigate a reconciliation difference that smells wrong rather than small — the size of a variance isn't always the size of the problem.
• Set the categorization rules in the first place, and revise them when the business changes.

The split is straightforward: AI carries the volume, the bookkeeper carries the judgment, and the platform keeps an append-only record of every categorization, state change, and reconciliation decision in between. That means the bookkeeper isn't defending the books from memory three quarters later — every action has a timestamp, an actor, and a prior state to point to. The work that used to consume the week becomes a morning review queue, and the hours that were going to data entry go to the calls and decisions that actually need a person.

Where to next

If your books, reconciliation, and close cycle no longer fit a generic subscription, the next step is to size what a custom build would look like for your business. Run the pricing calculator for a live cost range, see how the AI business assistant queries your books without hallucinating numbers, or review the infrastructure behind it. Bring the output to your CPA.